P2P Network for Very Large Virtual Environment Proceedings of the ACM symposium on virtual reality software and technology VRST '06.

P2P Network for Very Large Virtual Environment

Proceedings of the ACM symposium on virtual reality software and technology VRST '06

Adaptive Computing and Networking Lab

2

Outline

Introduction Overview System Structure Hierarchical LOD Object Structure Data Exchange Methods Simulation Conclusion


Introduction A centralized architecture is obviously not

a good framework to build a truly scalable Virtual Environment (VE).

P2P network overlays are potentially proper architecture.


Introduction This paper describes a peer connectivity

method for visualizing very vast and complex environments and sharing these environments among many users.



Key Concept

Neighbor peers in a VE are likely to have a lot of data in common

For example, if two peers have nearly the same viewpoint in a city model, data stored in both peers will be almost the same.


Challenges

Finding and maintaining the appropriate peer connectivity is a very difficult problem in a changing environment where peers viewpoint are allowed to move freely and peers can disconnect or appear at any time.



Peer Connectivity Method

Self-Adaptation:the ability for a requesting peer to dynamically take into account the serving capacity of neighbor peers



Three types of peers：Memory Peers (MP)

Likened to central servers

Visualization Peers (VP) End-user peers

Connectivity Peers (CP) Assigned the task of achieving the connectivity bet

ween peers



Each CP manages all VPs connected to it and, upon a request of a connected VP, it provides a list of the N VP neighbors, N being a parameter of the request.

The list of VP neighbors is the allocated serving pool (ASP).




Goal:to distribute the data requests of a VP among its neighbor VPs.



The selection of the serving VPs is made according to the following characterizes:Time To Server (TTS).The estimated available dataThe upload limit in bandwidth (BPUPL) assigned

to this peer for P2P serving tasks.The number of peers currently served by the

serving peer (NBPCS).


Simulation Accuracy:

ACC = NMISS / NREQ

Accuracy at time t is measured as the number of missing nodes (NMISS) in the level-of-detail (LOD) trees with respect to the number of required nodes (NREQ) to render the current viewpoint.


Simulation


Progressive Building Tree: PBTree


LOD Description Tree


Using LODDT to render different viewpoints


Using LODDT to assess the content of neighbor peer cache


Data Exchange Methods


Data Exchange Methods

Four types data exchange methods：From the MP only (C/S)From the peer having the least TTSFrom the closest peerUsing self-adaptive distribution method.


Simulation


Simulation


Simulation


Conclusion This paper proposes a new spatially-organized P

2P network overlay with intent to obtain self-scalability for a truly scalable Virtual Environment (VE).

It takes into account peer proximity along with some parameters and state variables that allow to assess the current serving capacity of the neighbor peers.

P2P Network for Very Large Virtual Environment Proceedings of the ACM symposium on virtual reality software and technology VRST '06.

Documents

networking lab slide

adaptive computing

networking lab introduction

networking lab challenges

pbtree slide

number of peers

peers viewpoint

appropriate peer connectivity